Molecular Vision 2006; 12:478-484 <>
Received 6 September 2005 | Accepted 6 May 2006 | Published 11 May 2006

Proteomic analysis of soluble factors secreted by limbal fibroblasts

Shigeto Shimmura,1,2 Hideyuki Miyashita,1 Kazunari Higa,2 Satoru Yoshida,2 Jun Shimazaki,2 Kazuo Tsubota1,2

1Department of Ophthalmology, Keio University School of Medicine, Tokyo, Japan; 2Department of Ophthalmology and Cornea Center, Tokyo Dental College, Chiba, Japan

Correspondence to: Shigeto Shimmura, MD, Department of Ophthalmology, Keio University School of Medicine, 35 Shinanomachi, Shinjuku, Tokyo 160-8582, Japan; Phone: +81-3-3353-1211; FAX: +81-3-3359-8302; email:


Purpose: To identify soluble factors selectively secreted by limbal fibroblasts as possible regulators of limbal basal epithelium.

Methods: Limbal, corneal, and conjunctival fibroblasts were first expanded in vitro in Dulbecco's modified Eagle medium containing 10% fetal bovine serum, and then maintained in serum-free medium for two weeks. Proteomic analysis of culture supernatants was done to compare differences in secreted matricellular proteins. Real time PCR and western blots were done to confirm the expression of secreted protein acid and rich in cysteine (SPARC), a protein found in abundance in extracellular proteins secreted by limbal fibroblasts. Immunohistochemistry of SPARC was done in human limbal tissue to show the spatial distribution of the protein. An adhesion assay was designed to demonstrate the effects of SPARC on an SV40 immortalized human corneal epithelial cell line (HCEC).

Results: Proteomic analysis revealed several proteins selectively secreted by limbal fibroblasts. The particular spots were identified as SPARC, vimentin, serine protease, collagen alpha 2 precursor, tissue inhibitor of metalloproteinase 2 (TIMP-2), and 5,10-methlenetetrahdrofolate reductase (FADH2). The expression of SPARC was confirmed by western blot analysis, and mRNA expression was significantly higher in limbal fibroblasts compared to central corneal fibroblasts when analyzed by real time PCR. Immunohistochemistry revealed higher distribution of SPARC in the subepithelial stroma of the limbus compared to the central cornea. The addition of 10 μg/ml murine SPARC in HCEC significantly reduced cell spreading at three h.

Conclusions: The matricellular protein SPARC is preferentially secreted by limbal fibroblasts, and may modulate intercellular adhesion of basal limbal epithelial cells.


The limbal basal epithelium has distinct characteristics compared with the corneal epithelium in the expression of several genes including increased α-integrin, ATP binding cassette protein 2 (ABCG2), and decreased keratin 3 (K3) and connexin 43 [1]. The differential expression of these markers is often raised as evidence for the presence of stem cells in the basal limbal epithelium. Accumulating evidence from clinical studies also support the limbal stem cell hypothesis, with successful ocular surface reconstruction reported by several laboratories following limbal transplantation [2,3], and more recently, cultured limbal epithelial sheet transplants [4,5]. The stromal niche is believed to modulate the phenotype of overlying epithelium, which probably involves soluble factors as well as regulation by direct contact. The plasticity of epithelial cells according to the underlying stroma was demonstrated by the use of amniotic membranes, and also by reversing epithelium/stroma combinations [6].

In order to screen for differences in secreted proteins by limbal and corneal fibroblasts, we performed 2-D PAGE (proteomic analysis) of condensed supernatants of serum-free cultured cells. After six proteins were identified, we further analyzed the distribution and function of secreted protein acidic and rich in cysteine (SPARC), also known as osteonectin. SPARC is a 43 kDa protein that contains a COOH-terminal extracellular (EC) module with two Ca2-binding domains and a follistatin-like module shared by a family of SPARC-related genes [7]. SPARC is also expressed by corneal epithelial cells, and is believed to be involved in the wound healing process of both the epithelium and stroma of the cornea [8,9]. In addition, SPARC modulates cell growth and adhesion of vascular endothelial cells [10], and has been reported to promote cancer cell migration and invasion [11]. Several of these physiological functions reported in SPARC are consistent with properties expected of soluble factors in the stromal niche of the limbus. Epithelial cell precursors are believed to be less dependent on intercellular communication, which in turn maintain these cells in an undifferentiated state. Evidence for this is given by the limited expression of the gap junctional protein, connexin 43, in basal limbal epithelial cells [1]. We therefore hypothesized that SPARC secreted constitutively by limbal fibroblasts can regulate epithelial cell adhesion.



Mouse recombinant SPARC and fibronectin were purchased from Sigma-Aldrich (St. Louis, Mo). Chemicals for proteomic analysis were obtained from Wako Pure Chemical Industries (Osaka, Japan) unless otherwise noted. The SV40 transformed immortalized human corneal epithelial cell line (HCEC) was a kind gift from Dr. Kaoru Araki-Sasaki (Kagoshima Miyata Eye Clinic, 1-5-1, Nishida, Kagoshima, Japan) [12].

Cell culture and adhesion assay

Human donor corneas not suitable for transplantation were obtained from the Northwest Lions Eye Bank. The epithelium and endothelium were bluntly removed with a gill knife, and stromal tissues were cut into small segments (approximately 2 mm x 2 mm) to allow fibroblasts to migrated during culture. Fibroblasts were cultured in Dulbecco's modified Eagle medium (DMEM) medium containing 10% fetal bovine serum until confluent, and then in serum-free DMEM for two weeks prior to proteomic analysis. HCECs were maintained in supplementary hormonal epithelial medium (SHEM), a 1:1 mixture of DMEM and Ham's F12 medium (DMEM/F12; Gibco, Invitrogen Corporation, Carlsbad, CA) containing 15% fetal bovine serum, insulin (5 μg/ml; Sigma-Aldrich, St. Louis, MO), cholera toxin (0.1 μg/ml; EMD Biosciences, San Diego, CA), human recombinant epidermal growth factor (10 ng/ml; Gibco), dimethyl sulfoxide (0.5%; Sigma-Aldrich), penicillin (0.7 mg/ml; Wako Pure Chemical Industries), and streptomycin (1.39 mg/ml; Wako Pure Chemical Industries).

Proteomic analysis

Supernatants from limbal, central corneal, and conjunctival fibroblasts cultured in serum-free DMEM for two weeks were collected by centrifugation. In brief, supernatants were placed in ultrafiltration tubes (Vivaspin 20; Sartorius, Goettingen, Germany) and centrifuged (MX-300; Tomy Seiko Co., Tokyo Japan) to remove proteins with molecular weights of less than 3 kDa. Lysis buffer (8 M urea, 2% NP-40, 2% ampholine (pH3.5-10), 5% 2-ME, protease inhibitor) was then added to the supernatant, and centrifugation was repeated. Two-dimensional PAGE was performed as previously described in the literature [13]. In brief, the first dimension was based on isoelectric focusing (pH 3.5-10) using a disk gel (Nihon Eido, Tokyo, Japan), followed by the second dimension done by SDS-PAGE in a 16.8% acrylamide gel (Bio-Rad Laboratories, Hercules, CA). Protein spots were visualized by Coomassie brilliant blue (CBB). Selected spots were dissected and digested with trypsin in 0.1 M ammonium hydrogen carbonate containing 10% acetonitrile for 16 h at 37 °C. Peptides were extracted from the gels with 60% acetonitrile containing 0.1% trifluoroacetic acid and then vortexed for 30 min. Peptide fragments were separated by C18 column (Magic C18 P/N 902-61260-00; AMR Inc., Tokyo, Japan) in a linear gradient (5-60%) of acetonitrile containing 0.1% formic acid. Separated peptides were analyzed by ion-trap mass spectrometry (MS, LCQ DECA; Thermoquest Corp., San Jose, CA) using a nanospray ionization apparatus. MS data analysis was done using Sequest (Thermoquest) and the Mascot Internet version [14].

Western blot

Western blot was used to confirm the expression of SPARC by cultured limbal and corneal fibroblasts, as well as primary cultured corneal epithelial cells. Culture supernatants were collected after two weeks of culture in DMEM containing 10% FBS and stored without condensation for western blot analysis. Cell pellets were dissolved with lysis buffer (50 mM Tris-HCl, pH7.4, 150 mM NaCl, 1% Nonidet P-40) and homogenized. Samples were incubated for 40 min at 4 °C, and then centrifuged at 15,000 rpm for 30 min at 4 °C. Protein concentration of the supernatant was determined by the DC protein assay (Bio-Rad Lab). All samples were then diluted in 2X sample buffer (100 mM Tris-HCl, pH 6.8, 4% SDS; Gibco, Invitrogen, Carlsbad, CA), 20% Glycerol (Wako), 12% 2-mercaptoethanol (Wako) and boiled. Ten μg of each sample (5 μg for β-actin) were loaded on a Novex NuPAGE 10% Bis-Tris gel (Invitrogen) and transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore, Billerica, MA). Membranes were blocked with 5% skim milk (Difco Laboratories, Detroit, MI) and 1.5% normal donkey serum in PBS for 60 min at room temperature. Membranes were reacted with an anti-SPARC antibody (1.B.789; US Biological, Swampscott, MA) for 60 min at room temperature. After three washes in TBST, donkey biotinylated antimouse IgG (Jackson ImmunoResearch) was added for 30 min at room temperature. Protein bands were visualized by the Vectastain ABC Elite Kit (Vector Laboratories, Bulingame, CA) using DAB (Vector Laboratories) as a substrate.

Real-time polymerase chain reaction

Total RNA was isolated from cultured limbal and corneal fibroblasts using the SV total RNA isolation system (Promega Co., Madison, WI) according to the manufacturer's recommendations. cDNA was prepared from total RNA with oligo (dT) priming and AVM reverse transcriptase XL (Takara, Bio Inc., Shiga, Japan) by incubation of a 25 μl mixture at 41 °C for 1 h. cDNA was subjected to PCR using the gene specific oligonucleotide primers and probe (5'-ACC CCA TTG ACG GGT ACC TCT CCC A-3'). Real-time reverse transcriptase polymerase chain reaction (real-time RT-PCR) using TaqMan chemistry (Applied Biosystems, Foster City, CA) and the ABI Prism 7700 Sequence Detection System (Applied Biosystems) was used to semiquantitate SPARC expression in limbal, corneal, and conjunctival fibroblasts. PCR products were detected during the exponential phase of the reaction in order to semiquantitate SPARC expression by each cell type (n=3).


Frozen sections prepared from a donor human cornea embedded in 4% carboxymethyl cellulose (CMC; Finetec Co., Ltd., Japan) were fixed in 4% paraformaldehyde (PFA) for 10 min. The fixed sections were treated with liberate antibody binding solution (L.A.B.; Polyscience, Inc., Warrington, PA.) at room temperature for 15 min for antigen retrieval. Antibodies used were antiosteonectin (Haematologic Technologies, Inc. Essex Junction, VT) and Cy3-labeled antimouse IgG secondary antibody. Isotype rat IgG (Chemicon) was used as control. The sections were incubated with 1 μg/ml 4',6-diamidino-2-phenylindole (DAPI; Dojindo Laboratories, Tokyo, Japan) at room temperature for 5 min. Finally, sections were washed three times in Tris-buffered saline tween (TBST) and coverslipped using an antifading mounting medium (50 mM Tris buffer saline, 90% glycerin; Wako), and 10% 1,4-diazabicyclo (2,2,2) octane (Wako).

Adhesion assay

One of the established physiological effects of SPARC is the suppression of vascular endothelial cell growth and adhesion [10]. In order to pursue the possibility that SPARC may have similar effects on corneal epithelial cells in vitro, we performed a modified version of a cell adhesion study reported previously [15]. Nontreated 96 well plates (260887, Nalge Nunc Int, Rochester, NY) were coated with 100 μl of fibronectin in phosphate-buffered saline plus (PBS+; 1 μg/ml) at 4 °C overnight, and washed with PBS. Serum-free DMEM with or without murine SPARC (final 10 μg/ml) were added to the wells. HCEC were trypsinized, neutralized, resuspended in serum-free DMEM, and a 50 μl sample was added to each well (104/well). After 3 h incubation at 37 °C, the central area of each well (856 μm x 678 μm) was photographed using the Axiovert 200 microscope (x10, Carl Zeiss, Gottingen, Germany). Cells were scored as previously described [16]. Round cells with no apparent signs of spreading were given a score of 3. Rounded cells with short cellular processes were assigned a score of 2. Spread, flattened cells were given a score of 1. Adhesion score for each well was calculated by the average score of all visible cells in a randomly selected field of view.


Proteomic analysis

2-D PAGE of supernatant from limbal fibroblasts is shown in Figure 1. Total protein levels were low, in general, since this was an analysis of culture supernatants and not of homogenized cells. Although samples were condensed prior to electrophoresis, only blots that were dense enough to allow sequence analysis were further investigated. The six proteins specifically identified in the supernatant of limbal fibroblasts along with their accession numbers are listed in Table 1.

Constitutive expression of SPARC by limbal fibroblasts

We further pursued the possible role of SPARC as a major matricellular protein in the limbal stroma. Real-time PCR was done to semiquantitate SPARC mRNA transcription in cultured cells, and the result was consistent with the higher protein content in limbal fibroblasts observed in the proteomic analysis (Figure 2). Western blot results confirmed SPARC protein secreted in the supernatant of limbal and corneal fibroblasts (Figure 3). SPARC was also expressed by primary corneal epithelial cells, however, the expression levels were lower compared to limbal and corneal fibroblasts.


The cumulative data show higher expression of SPARC in limbal fibroblasts in vitro, but does not necessarily reflect that this applies in vivo. We therefore performed immunohistology using an anti-SPARC monoclonal antibody in fresh donor limbal tissue to observe the distribution of SPARC in situ. As shown in Figure 4A, a higher level of SPARC-associated Cy-3 fluorescence was observed in the subepithelial regions of the limbus compared with the central cornea. The difference can be appreciated when compared with the uniform fluorescence observed in the overlying epithelial cells (Figure 4B). Thus, SPARC is constitutively expressed in the limbal stroma by resident fibroblasts without the stimulation of a wound healing process.

Cell adhesion assay

In order to observe the effects of SPARC on corneal epithelial cells in vitro, an immortalized cell line (HCEC) were used to observe for changes in cell adhesion and morphology. The addition of SPARC in the culture supernatant resulted in rounding of individual HCEC after 3 h (Figure 5). The difference was statistically significant using a rounding index originally described by Lane and Sage [16] (n=5).


SPARC is a 43 kDa protein that contains a COOH-terminal EC module with two Ca2+-binding domains, a follistatinlike module, and an NH2-terminal acidic module [7]. The expression of SPARC by corneal stromal cells has been reported to play a role in the wound healing response, evidenced by the upregulation of SPARC by the fibroblast and myofibroblast phenotype [9]. However, SPARC was not detected in quiescent corneal stromal cells, and hence, the major function of the protein was speculated to be related to wound healing. Conversely, SPARC secreted by epithelial cells was shown to induce contraction of stromal fibroblasts in vitro, suggesting that SPARC is a key protein in epithelial/stromal interaction of the cornea [17]. SPARC has also been proposed to be involved in corneal epithelial migration and stratification following mechanical ablation [8].

We found that limbal fibroblasts secreted higher levels of SPARC compared to central corneal fibroblasts in vitro without stimulation by serum or cytokines, and also in vivo without any wound-healing stimuli. SPARC was one of only a few proteins detectable by proteomic analysis in the limbal fibroblast supernatant, suggesting a functional role in the homeostasis of the limbal structure. Although it can be argued that corneal fibroblasts cultured in vitro are not the same as keratocytes in vivo, experiments requiring large quantities of cells would not be possible without in vitro expansion. All cells used in the current study were first expanded in vitro using serum containing 10% serum, therefore, the phenotype of these cells at the time of analysis is not necessarily consistent with the normal phenotype. The results of preotomics alone, therefore, have limits without further analysis. Interestingly, one of the proteins detected in the limbal cell supernatant was vimentin, an intracellular intermediate fiber, suggesting that some of the proteins in the supernatant may have been the result of apoptosis. We did not pursue this issue further, and chose to focus on SPARC which is a secreted protien.

Figure 4 shows the immunohistochemistry of SPARC in donor cornea tissue. The result shows that SPARC is expressed more in the limbal stroma, reflecting the results of real time RT-PCR and western blots of cell supernatants. We have also found through immunohistochemistry that epithelial cells were positive for SPARC, while western blots only detected trace levels of SPARC from cell lystates and supernatant of primary cultured epithelial cells. This may be explained by the fact that SPARC is secreted by corneal epithelial cells during wound repair [8,17], and that primary cells in vitro may be in a state similar to epithelial cells undergoing wound healing.

We focused on the function of SPARC, since the matricellular protein has been reported to regulate the adhesion of bovine aortic endothelial cells [10]. Using a previously described adhesion assay, we found that exogenous SPARC inhibited adhesion of a human corneal epithelial cell line, which may be due to the Ca2+-binding ability of SPARC. Espana et al. [6] have previously reported that limbal stroma, and not corneal stroma, was required to maintain an undifferentiated phenotype (K3 negative, Cx 43-low) in corneal epithelial cell sheets. This implies that soluble factors expressed by limbal fibroblasts may be involved in this phenomenon. The extracellular matrix and basement membrane components of the limbal area are distinct from the central cornea, as reported by several studies [18,19]. SPARC is also involved in the migration and invasion of prostate cancer cells [11] and breast cancer cells [20] through the activation of matrix metalloproteinase 2 (MMP2). These are several properties that are expected of matricellular proteins in the putative limbal stem cell niche. Interestingly, the MMP2-specific inhibitor, TIMP2 was also preferentially detected in the supernatant of limbal fibroblasts, suggesting that an intricate network based on a balance of effectors and inhibitors may be involved in the homeostasis of the limbal stem cell niche.

Growth factors such as keratinocyte growth factor and hepatocyte growth factor are also mediators of fibroblast/epithelial interaction involved in epithelial proliferation and migration [21]. Although the network of epithelial/mesenchymal interaction in the corneal limbus is sure to involve a wide variety of matricellular proteins, cytokines, and growth factors, the inhibition of cellular adhesion and cell/cell interaction by SPARC may be a major component of the limbal microenvironment. We found that the human amniotic membrane (AM) also contains SPARC (data not shown), which may partially explain the ability of AM to preserve the undifferentiated state of limbal epithelial cell in vitro [22]. While further studies are required to elucidate the interactions of soluble factors involved in the limbal niche, a combination of such components may be used to enrich limbal stem cells in vitro.


This study was partly supported by a grant of Advanced and Innovational Research Program in Life Sciences from the Ministry of Education, Culture, Sports, Science and Technology of Japan.


1. Chen Z, de Paiva CS, Luo L, Kretzer FL, Pflugfelder SC, Li DQ. Characterization of putative stem cell phenotype in human limbal epithelia. Stem Cells 2004; 22:355-66.

2. Kenyon KR, Tseng SC. Limbal autograft transplantation for ocular surface disorders. Ophthalmology 1989; 96:709-22; discussion722-3.

3. Tsubota K, Satake Y, Kaido M, Shinozaki N, Shimmura S, Bissen-Miyajima H, Shimazaki J. Treatment of severe ocular-surface disorders with corneal epithelial stem-cell transplantation. N Engl J Med 1999; 340:1697-703.

4. Koizumi N, Inatomi T, Suzuki T, Sotozono C, Kinoshita S. Cultivated corneal epithelial transplantation for ocular surface reconstruction in acute phase of Stevens-Johnson syndrome. Arch Ophthalmol 2001; 119:298-300.

5. Pellegrini G, Traverso CE, Franzi AT, Zingirian M, Cancedda R, De Luca M. Long-term restoration of damaged corneal surfaces with autologous cultivated corneal epithelium. Lancet 1997; 349:990-3.

6. Espana EM, Kawakita T, Romano A, Di Pascuale M, Smiddy R, Liu CY, Tseng SC. Stromal niche controls the plasticity of limbal and corneal epithelial differentiation in a rabbit model of recombined tissue. Invest Ophthalmol Vis Sci 2003; 44:5130-5.

7. Brekken RA, Sage EH. SPARC, a matricellular protein: at the crossroads of cell-matrix communication. Matrix Biol 2001; 19:816-27.

8. Latvala T, Puolakkainen P, Vesaluoma M, Tervo T. Distribution of SPARC protein (osteonectin) in normal and wounded feline cornea. Exp Eye Res 1996; 63:579-84.

9. Berryhill BL, Kane B, Stramer BM, Fini ME, Hassell JR. Increased SPARC accumulation during corneal repair. Exp Eye Res 2003; 77:85-92. Erratum in: Exp Eye Res 2003; 77:643.

10. Sweetwyne MT, Brekken RA, Workman G, Bradshaw AD, Carbon J, Siadak AW, Murri C, Sage EH. Functional analysis of the matricellular protein SPARC with novel monoclonal antibodies. J Histochem Cytochem 2004; 52:723-33.

11. Jacob K, Webber M, Benayahu D, Kleinman HK. Osteonectin promotes prostate cancer cell migration and invasion: a possible mechanism for metastasis to bone. Cancer Res 1999; 59:4453-7.

12. Araki-Sasaki K, Ohashi Y, Sasabe T, Hayashi K, Watanabe H, Tano Y, Handa H. An SV40-immortalized human corneal epithelial cell line and its characterization. Invest Ophthalmol Vis Sci 1995; 36:614-21.

13. Kajiwara H, Kaneko T, Ishizaka M, Tajima S, Kouchi H. Protein profile of symbiotic bacteria Mesorhizobium loti MAFF303099 in mid-growth phase. Biosci Biotechnol Biochem 2003; 67:2668-73.

14. Perkins DN, Pappin DJ, Creasy DM, Cottrell JS. Probability-based protein identification by searching sequence databases using mass spectrometry data. Electrophoresis 1999; 20:3551-67.

15. Girard JP, Springer TA. Modulation of endothelial cell adhesion by hevin, an acidic protein associated with high endothelial venules. J Biol Chem 1996; 271:4511-7.

16. Lane TF, Sage EH. Functional mapping of SPARC: peptides from two distinct Ca+(+)-binding sites modulate cell shape. J Cell Biol 1990; 111:3065-76.

17. Mishima H, Hibino T, Hara H, Murakami J, Otori T. SPARC from corneal epithelial cells modulates collagen contraction by keratocytes. Invest Ophthalmol Vis Sci 1998; 39:2547-53.

18. Ljubimov AV, Burgeson RE, Butkowski RJ, Michael AF, Sun TT, Kenney MC. Human corneal basement membrane heterogeneity: topographical differences in the expression of type IV collagen and laminin isoforms. Lab Invest 1995; 72:461-73.

19. Fukuda K, Chikama T, Nakamura M, Nishida T. Differential distribution of subchains of the basement membrane components type IV collagen and laminin among the amniotic membrane, cornea, and conjunctiva. Cornea 1999; 18:73-9.

20. Gilles C, Bassuk JA, Pulyaeva H, Sage EH, Foidart JM, Thompson EW. SPARC/osteonectin induces matrix metalloproteinase 2 activation in human breast cancer cell lines. Cancer Res 1998; 58:5529-36.

21. Li DQ, Tseng SC. Three patterns of cytokine expression potentially involved in epithelial-fibroblast interactions of human ocular surface. J Cell Physiol 1995; 163:61-79.

22. Grueterich M, Espana EM, Tseng SC. Ex vivo expansion of limbal epithelial stem cells: amniotic membrane serving as a stem cell niche. Surv Ophthalmol 2003; 48:631-46.

Shimmura, Mol Vis 2006; 12:478-484 <>
©2006 Molecular Vision <>
ISSN 1090-0535